Calculating Speakers: Codeswitching in a Rational Choice Model Author(s): Carol Myers-Scotton and Agnes Bolonyai Reviewed work

(s): Source: Language in Society, Vol. 30, No. 1 (Mar., 2001), pp. 1-28 Published by: Cambridge University Press Stable URL: . Accessed: 19/11/2011 13:53
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact

Cambridge University Press is collaborating with JSTOR to digitize, preserve and extend access to Language in Society.

Language in Society 30, 1-28. Printedin the United States of America

Calculating speakers: Codeswitching in a rational choice model
CAROL MYERS-SCOTTON Linguistics Program Universityof South Carolina Columbia,SC 29208 AGNES BOLONYAI

Departmentof English East Carolina University Greenville,NC 27858

Although the methodologies for describing many types of linguistic variation have been well developed, satisfactory theoretical links between data and explanation- especially links that include causal mechanisms- remain lacking. This article argues,somewhatparadoxically,thateven though most choices reflect some societal pattern, speakers make linguistic choices as individuals. That is, choices ultimately lie with the individual and are rationally based. RationalChoice Models (e.g. Elster 1979, 1989, 1997) provide explanatorymechanisms for the ways actors in society select from alternative structuresand available options. The Rational Choice approachtaken here is enhanced by diverse theories of human action (e.g. Damasio 1996, Klein 1998, Lessig 1995). Analysis of codeswitching examples within a recasting of the MarkednessModel (Myers-Scotton, e.g. 1993, 1998) suggests how a rationallybased model offers better explanations for linguistic variationthando other approaches.(Codeswitching, cognitive calculations, linguistic variation,MarkednessModel, Rational Choice)*

What is the engine that drives speakers to select one linguistic variety over another?Why speak English ratherthan Spanishwhile discussing a deadline with a fellow workerif you're both Chicanasin Los Angeles? Why switch to an approximation of a "good ole boy" Carolinaaccent when asking your car mechanic for advice in urbanSouth Carolina?That speakersalso choose one lexico-syntactic permutationof a routine, such as a directive or an apology, over anotheris less obvious, but this selection is equally in need of an answer.
C 2001 CambridgeUniversity Press 0047-4045/01 $9.50

Brown & Levinson 1987 for a related approach). Labov (1989:52. "Individualbehaviorcan be understoodonly as a reflection of the grammarof the speech community. specifically in recognizing that certain individual women.has stressed the correlation of individual choices with factors that delineate social group memberships and/or other features of the community-basedsocial context. However. we argue that talk follows principles of well as many other ingredientsthat define speakers "externally. what ultimately sets linguistic choices in motion is speaker intentions and calculations to optimize rewards. cited in Hudson 1996:30) writes. within one social groupthat is in conflict with another). Such an approachlocates the crucial impetus for linguistic choices within the largercommunity(or.they are more concernedwith the ways in which inferential processes. Probablythe most widely held view of the forces behind linguistic choices is the approacharticulatedin the 1960s by such figures as Labov 1966 andFishman 1965. following principles or maxims. but cited by many as affecting linguistic choices.but of the community.especially those outside the field ." such as gender and ethnic group. Further.Languageis not the propertyof the individual. but not women as a group. 2 Language in Society 30:1 (2001) .researchthat is synonymous with sociolinguistics for many. but ratherindividuals. Nevertheless. 1989). while of course their subject is just as much language use in social relations as is the subject in this paper.we acknowledge that factors not emphasized here. On this view. how utterances convey inferences about the speaker's negotiation of persona or other interpersonal relationships(cf.are innovatorsin a given community(Labov 1998). not the speech community nor even the social network. However.These principles hold that choices in specific interactionsare best explained as cognitively based calculations that depend on the actor's estimation of what actions offer him/her the greatest utility (Elster 1979. in some cases. enrich the referentialmessage of a linguistically encoded logical form.we will seek to demonstratethe primacyof a specific cluster of factors at the point when linguistic choices are actually made. necessarily "own" the linguistic choice of one way of speaking over another.thatis. explicating all contributingfactors in detail is not our goal. The argumentis this: by analogy to the switch that engages any engine. Therefore.and we specifically consider codeswitching data to supportthat point.To this end. and still generally embracedtoday."Labov recently has distanced himself somewhat from this view.CAROL MYERS-SCOTTON & AGNES BOLONYAI We first acknowledge that the judicious answer to these questions is that an engine has many parts. These factors include socio-economic class in particular.2We refer to this as the "social factors" approach. also have a place in the process.We areoccupied morewith the ways in which similar inferentialprocesses enrichthe SOCIAL MESSAGE .For example. the researchof Labov and his associates . in this article we support a Rational Choice Model to explain why speakers select one linguistic form over another. rather. in much the same way that Grice 19751 and Sperber & Wilson 1995 see it. Choosing one's words is seen here as purposive behavior.

the implicit claim in the Labovian approachhas been that if you have information about speakers in relation to specified variables (aspects of their social identities and/or the situation) from a sample of subjects across a community. in studying speakersas sometime "performers. However. beliefs and action opportunitiesgenerate a specific action. First. it is hardto see how such factors . Although we agree that aspects of the larger societal backgroundcertainly affect choices. Popper 1959). we propose an approachwhich proceeds from first constructing an analytical model. In contrast to a variable-basedanalysis whose main achievement is linking one independentvariableto another. Bing & Freed 1996). In seeking to explain variationin the speech of members of the same ruralcommunity. not explaining. For example. That is. to the analyst. linguistic choices primarilyreflect the speaker's place in a social group defined by the variable(s) studied.after years of sociolinguists' viewing gender as a unitary category. Hedstrom & Swedberg (1998:23) write. it is an unobserved construct useful for narrowingthe gap between cause and effect. and no descriptionis theory-free (cf.CALCULATING SPEAKERS Researchwithin this paradigmemphasizes the quantitativedescriptionof observable behavior.even if a statistically significant numberof members of a groupmake the same choice in a given context. many researcherswithin a variationist frameworknow seem to realize that explanationsfor individual variation may not lie exclusively with demographicvariables. it simply summarizesthem.3At the same time. and attitudes and values than conventional social divisions. variable-basedanalysis rarely can even delineate variables affecting all choices."In addition. choices. In their introductionto a collection of essays on social mechanisms. correlationimplies causation. Eckert 1989." To the actor.the emphasis is still on describing." Schilling-Estes (1998:68-9) interpretsstyleshifting as "primarilya means whereby speakers alter the images of self which they projectfor others. we argue that the variation in such factors DOES NOT DIRECTLY DETERMINE actual choices. most of today's researcherson language and gender recognize that gender on its own does not account for the choices thatwomen make (cf. "This type of mechanism [an action-formationmechanism] shows how a specific combinationof individual desires. for two reasons. Even though this typically includes pointing out patternsin that group. interactionalrelations. no studies show thatALL members do this. with a social mechanism at its center. because a variable-centeredanalysis does not explain decisions. 1999. a social mechanism is an abstractdevice useful for guiding decisions. "Ourdata suggest that some variationfor the regional accommodationstructuresmay be a function more of personal history. Thus. "A mechanism-basedexplanation seeks to provide a fine-grained as well as tight coupling between EXPLANS Language in Society 30:1 (2001) 3 . or even constructedsocial identities. Wolfram and Beckett (1999:24) conclude. Bergvall. Further.can be construedas a mechanism for making choices. Thus. for example ."not necessarily as a productof the context. you are essentially explaining what drives their choices.dependentvariable (such as linguistic variation).

"4 Although CA was originally developed to analyze Elster's terms.especially the natureof certain adjacencypairs. We employ input from social theories such as that of social mechanisms to enhance the base of sociolinguistic theory. can be considered devices thatconstrainspeakersto view certainpotentialchoices as preferredand others as not. social factors and other institutionalconstraintsdeterminethe speaker's "opportunityset. As an explanationof choices.we heartily agree that structuralfeatures of any conversation. that social factors (or specific ethnographicmilieux) do not impinge on choices. Another approachto dealing with linguistic choices that has many adherents today is ConversationAnalysis (e. Of course. "Instead. as well as the interaction-ascontext and its dynamicquality.Generally." Further. the structuralfeatures studied by CA offer an exceedingly "flat"explanationof choice.CA practitioners have a very different view of context from thattakenby sociolinguists within the social factors framework.CAROL MYERS-SCOTTON & AGNES BOLONYAI and EXPLANANDUM" (Hedstrom & Swedberg 1998:25). Although the model that we will employ to analyze data in this article is centeredon how actors make choices. the point is that they do not ordainthe actual choices that individual speakers make.g.Just as importantis the dual role of the speaker and immediate addressee in shaping the dialogue." In our terms.Withinstructure."They go on. reliance on surface structureas an explanationfor an individual'scontribution a conversationignores two sourcesof to 4 Language in Society 30:1 (2001) . the majorfocus is on sequencing.g. ConversationAnalysis (CA) offers the systematic propertiesof structural organization(e. recently it has been employed to interpretthe structureof bilingual interactions. the opportunityset is a speaker's linguistic repertoire. turn-taking. one would have to posit that situational incentives and speakers'motivationsmust be the same at all times and places. as is RationalChoice Theoryin general.In arguingagainst such a posito tion. Drew & Heritage 1992). adjacencypairs) as what speakersorientto in makingconversationalcontributions.This point is developed below.5 However. That is. the CA perspectiveembodies a dynamicapproachin which 'context' is treatedas both the project and product of the participants'own actions and therefore as inherentlylocally producedand transformableat any moment.ourmodel also explicitly recognizes thata currentspeaker'schoice dependson the addressee'sresponse (on manydifferentlevels) for its "success.g. Duranti& Goodwin 1992. A second problemwith the idea that linguistic variationis explained through variation in social factors is that it implies that OTHER factors. Rather. especially those "internal" the speaker. Drew & Heritage (1992:19) explicitly reject "whatmay be termedthe 'bucket'theory of context in which some preestablishedsocial framework is viewed as 'containing'the participants'actions.of course. For example.are static or irrelevant. they are not. especially codeswitching (e. Auer 1998). Elster (1979:115) makes the point that for "opportunities"(derived from situational factors) to explain choices. No one is arguing.

Who participants"are". is ironic network. thougha social factorsmodel can providegeneraloutlines to accountfor the majorityof choices. according to Hutchby & Wooffitt 1998:45).receives little attention. when its adherentstalk about the "local producto" tion"of identities/meaning.Further. another. including the content and form of the interlocutor'sturn. Milroy & Milroy 1995) and CA have offered useful descriptions of linguistic choices (cf.therefore. speakers"know"the interpersonalassociations that one form of a conversational sequence. In contrastto such models. As we will could arguethatthe speaker's"orientation what the hearerdoes is itself a rationalact. it ignores the "texture" vide to conversationalpartners.or even ignores speaker motivations. First.studyof ing).CALCULATING SPEAKERS that aspects of the wider social context prochoice.has been to demonstratethat there are indeed predictable macro-patternsand a hierarchy among the social identity factors associated with variationin the patterns. Second. we arguehere for a model based on assumptionsof preferences and intentions. specifically on codeswitchsociolinguistics.g. ( the extent that motivations are considered.andeven ethnographic key aspects of the tacit knowledge that speakershave developed thatCA neglects through their very conversations (cf. Duranti& Goodwin 1992 andAuer 1998. CA also downgrades.analyses in some recentCA-based studies leave more themselves to produceorrenderconsequentialthe roomfor individualparticipants salient social categories or conversationalelements (e.This knowledge includes not just HOW certain interactionsproceed (which CA does study). The majorcontribution Labovian-stylevariationist ing the language use patterns of speakers as members of groups . they are discussed as a feature of certain individuals at certain times. like the social factorsmodel. Labov 1990.g. rationalitymeans cognitively based calculations. but ALSOthe socio-psychological associations and. however.However.The claim dea veloped hereis thatthe engine drivinglinguisticcode choices is rationality. Goffman 1974 on "frames"). mechanismuniversallyavailableto humans.we recognize thatboth social factorsapproaches(cf. In summary.calls up.6There is too much of a gap between the macro-level entities of social factors models and the sequential organizationof CA and individual choices. certainly it has been crucial in making us aware of the intricate organizationof everyday conversation.That demographic. Briggs 1998). it cannot explain all the choices. operatingon perceived opportunities. neither approachoffers adequate explanationsfor individualvariationin linguistic choices orientedto the behavior of others. Language in Society 30:1 (2001) 5 .the social messages carriedby one linguistic choice ratherthan another.However. different renditions of an apology) vs. in addition to ways of speakingby which specific actions get acknowing the "institutionalized complished" (what CA is concerned with.When all is said and done.Speakersarerationalin the sense thattheir choices dependlargelyon assessmentsof possible optionsin termsof a cost-benefit analysis thattakes accountof theirown subjectivemotivationsandtheirobjective opportunities. but not as universally present.For example. in CA's favor. As for it terms. rationalbehaviorinvolves paying attentionto available evidence.Moreover.

RC Theory does not seek a correlation between language choice and activity type. This example consists of dinner-table conversationbetween Kristof(K).That is. and to minimize costs. All are very fluent in English (the fatherperhaps less so). The correlationin RC Theory is always between language choice and the agent's intentto act rationally. Given its premises and goals.To this end. certainly. or between language choice and the sequential natureof the conversation. Hungarianis the preferred language of family interactions. It does not attemptto accountfor the characterof social groups and theirrelations.The researchquestion addressedfor all examples is this: What motivates speakersto switch languages within a single conversation?The argumentto be supported. a boy eight and a half years old. and they wish to maintaintheir children's Hungarian. we analyze some examples of codeswitching between languages. we state the research question and illustrate the type of data to be considered. All data come from naturallyoccurringconversations.son. variationin choices . showing how an RC approachproduces a more principled analysis than do other possible approaches.Yet RC theorydoes not neglect large-scale factors.the dominantuse of Hungarianat the dinnertable.or for the natureof conversation. 1 illustratesthe natureof the data to be considered.Threeexamplesinvolving Hungarian/English were audio-recorded CS in the family of Agnes Bolonyai. A sketch of our theoretical approachfollows.arebest explainedby an analysis assuming that choices depend on the speaker'sestimation of what choices offer him or her the greatest benefit. and all are bilingual in Hungarian(their first language) and in English.we will analyze a numberof examples of codeswitching (CS) between two languages. it places choice at a point after these factors have had their effects.father. and daughterlives in the United States. The family of mother. English is the main language of his interactionswith everyone except his parents.It attemptsto explain what people do. GIVEN the fabric of society and social interaction. DATA TO BE STUDIED To illustrate the types of interpretations linguistic choices that follow from a of model based on RC theory. the second author.based on available evidence. That is.she is the "mother"in all three examples. as will become clear. specifically for these data but generally applicable. choices reflect a goal to enhance interpersonalrelations and/or material or psychological rewards. the parents speak Hungarianalmost exclusively. he 6 Language in Society 30:1 (2001) . is this: Linguistic choices and specifically. Next.CAROL MYERS-SCOTTON & AGNES BOLONYAI We emphasize what Rational Choice (RC) Theory does not do. Ex. In the next section. Between themselves. Although Krist6f accepts .and contributesto .7 The organization of this article is as follows. it is the parents' unmarked choice for dinner-tableconversations.The final section offers some conclusions. The other examples come from the publishedCS literature. instead. and his mother(M).

hozzd mdst is. We were drawnto the argumentsof anothersocial scientist.perhapsmore unlikely . topic) or sequentialorganization. English prevails with some switching to Hungarian. mi volt az iskoldban?Irtatoktesztet? 'Krist6f. we use a modified version of Elster's filter model of how RC theory conceptualizes decision-making.ideas and datafrom three other. GaryKlein (1998). how was school? Did you write any tests?' 2 K: Tesse'k? 'Pardon?' 3 M: Voltteszt? Milyen volt az AGP? 'Were there any tests? How was AGP?' 4 K: Most nem voltam. The amended Markedness Model is illustrated explicitly as part of the analysis of ex. It hasn't startedyet. people usually do what they believe is likely to have the best overall outcome. (1. 'I wasn't [in AGP] today. 'It's on Thursday. More specifically.0) llyen kicsi tdnyerokban 11 csindljdka restaurant-okban. (1) Making a salad at the dinner table (Hungarian/English). 13 M: Odaadtamaz egeszet. 1998) as more explicitly an RC model. 1 in this section. 1 M: Krist6f. too. Elster's views can be summarizedby this statement:"When faced with several courses of action. we recast the MarkednessModel of Myers-Scotton (1983. but it is also applied to other examples in a later section. even with his younger sister.' (K making the salad) 12 K: I need some salad. THEORETICAL APPROACHES In this article. Add something else to it. who has studiedsuccessful split-seconddecision-makingwhen Language in Society 30:1 (2001) 7 . 1. RC theorists come from a variety of disciplines. 1993.' saldtdt. our framingof the markednessmodel as an RC model largely follows the ideas of Elster. 9 M: Mi? 'What?' 10 K: I'll make my own salad.sources inform the analyses. why does Kristof use Hungarianto speak with his mother about school matters.butthenproducetwo turnsin English andone in English/Hungarianwhen he talks about food? His choices cannot be explained by situationalfactors (e.The example is analyzed in a following section.CALCULATING SPEAKERS speaks only English at school and with his friends.' In ex. Tegye'l 'I've given you all.g. wouldn't you?' 8 K: I'll make my own salad. please. ugy-e? 7 M: Ke'rtek 'Wouldyou like some salad.a social scientist who has writtenextensively aboutrationality. This deceptively simple sentence summarizesthe theory of rational choice" (Elster 1989:22). Nem volt meg. In addition. 'They make it in such small plates at the restaurants.' 5 M: Kedden van? 'Is it on Tuesday?' 6 K: Thursday-nvan.

1998). Myers-Scotton 1983. although it was said that markedchoices are negotiationsto change the social distance.g. the evaluatorindicates which choices are relatively more or less markedfor the interactiontype.Specifically as a sociolinguistic construct. (a) The MM presupposesthatas partof theirgeneralcognitive architecture all speakershave a MARKEDNESS EVALUATOR. speakers develop the ability to provide relevantINTERPRETATIONS for all choices. This abstractcomponentunderliesthe capacityto conceptualizemarkedness.CAROL MYERS-SCOTTON & AGNES BOLONYAI physical wellbeing is on the line .for example. but they did not develop precisely how linguistic choices translateinto social meanings. markednessrefersto the capacityto develop the following threeabilities. decisions by firefighters and militarypersonnel. as well as on how the individualinteractiondevelops. speakers learn to recognize that the markednessorderingof choices is dynamic. a law professor.For example. he discounts RC models). We also make indirectuse of the ideas of LawrenceLessig (1995:951). the model offered no principledbasis to arguefor one interpretation involving a change in the social distance that is not gained by another. Damasio provides empirical evidence about how organisms (including humans) develop "somaticmarkers" help limit the "space"necessary for decision-makingand that also allow the organism to call on previous experience to enable it to shortcut comparisonsin decision-makingfor survival.. or statuses. 1993." Finally. Myers-Scotton 1993) argued for the central role of cognitively based calculations. 1998) also figure in our analyses. andhis claim thatintuitiongrows out of experience (1998:33). (iii) Finally. (i) Most importantis the perception that relevant linguistic choices for a specific interaction type fall along a multidimensional continuum from more socially unmarked to more marked. within a particularcontext. One selects certainwords over otheracts in some contexts. nor did they emphasize the link with dependson the specific interaction type. "theyare also used. Thatis.Klein's approachdiffersfrom thatof ElsterandotherRational Choice proponents(in fact.who refers to what we might call the consequences of social norms as SOCIAL MEANINGS: "the semiotic content attachedto various actions. or inactions." so that "[social meanings] not only constitute." He goes on to say (1995:956) that social meaningsnot only exist. marked as well as unmarked. 8 Language in Society 30:1 (2001) . whether an individually or collectively chosen end . or constrain. abbreviatedsketch An of its premises follows (cf. 1996.. or guide. one chooses a certain languageto signal one meaningratherthananother.given the interactiontype. (ii) In addition.they are also tools means to a chosen end. Earlierformulationsof the MarkednessModel (e. (b) To develop these abilities requiresexposure to the use of both unmarked and markedchoices in actual community discourse. but its value for us is its emphasis on what Klein calls INTUITION.The MarkednessModel (MM) has been employed by many.but also criticized and misunderstood.the ideas of a neurobiologist (Damasio 1994. Inputfrom experience empowers the markednessevaluatorto set up readingsof markednessas a guide for speakersto prunemultiple options.

but rather a PROCESSfor evaluating potential choices. the tone of a job interview may move from relatively formal to decidedly informal if the participantsdiscover they are both from the same distant small town. this move would index a markedRO set from the point of view of family norms. any choice a speakermakes is perceived as indexing a desired All Rights and Obligations(RO) set between participants. participantsinterpreta choice against the backdropof those choices that index the more unmarkedRO sets for a specific interactiontype. What the markedness evaluator offers is NOT a set of rules. to speak only Hungarianat the dinnertable would be an index of what his parents might prefer as the unmarkedRO set for family interactions in his home. too. (f ) RO sets are the elements in this model that are directly derived from whatever societal factors are salient in the community and in the interaction type. Kristof would be asserting his independencefrom familial control and possibly even his "defection" to the dominant (American) culture. the choice of Krist6f.the RO set that codeswitching indexes may be the more unmarkedone in many immigrantfamilies. but at the same time.However. the children maintaintheir first language and thereforekeep theirethnicity salient. (d) The interpretationsthat speakers attach to linguistic choices have to do with the speaker's projection of his/her own persona and relations with other participants.That is. As a corollary. Under such an RO set. RO sets. (e) The markednessof an RO set for a specific interactionis open to change (based on change in situationalcomponentsor participants'negotiation). the markednessevaluator is a DEDUCTIVE device (it makes predictions about relative markedness)that operates on inductively assembled data. For example. children are compliant with their parents'wishes to keep theirethnicity salientthroughlinguistic means ratherthan assimilatingfully to the dominantculture. Unmarked/marked matches between choices. Although they depend on hierarchies of salience of features that preexist the interaction(i. Language in Society 30:1 (2001) 9 . this might index a somewhatless preferredRO set. if Krist6fshould insist on speakingonly English at the family dinnertable. no matterwhat is preferredby parents. given the salience of who the participantsare and of other relevant situationalfactors. the markednessof the currentlinguistic choice for the interaction changes. just because it representsa compromise. speaking boy For example. then. norms or social meanings).e. and interactiontypes will vary according to the speech community.CALCULATING SPEAKERS (c) In sum. Under this RO set. from the parents' point of view. (g) Unmarkedchoices (indexing an unmarkedRO set) are those that are more expected.and if it does change. a Hungarian-and-English living in the United States. they also depend on how the salience of these features evolves as the interactionunfolds.Thus.If Kristofswitches between Hungarian andEnglish.this means that they also recognize some choices as indexing more markedRO sets. they can index theiridentityin the widerAmericancommunity. Still.

manytypes of interactionsin institutionalsettings).or that between non-familiar supervisorsand subordinates. also from a foreman(Are you going to be here tomorrow?).Bernsten (1998:185-86) shows how one question directive from a foreman (Do you thinkyou can do Helen's job?) is taken as an insult. Because bald imperativesare very unmarkedin this work forty or fifty-something employer recalled a candidatewho greeted him as dude. (k) Not all speakers may view the more unmarkedchoice for a given interaction type as "fair. and how anotherone.CAROL MYERS-SCOTTON & AGNES BOLONYAI (h) For a variety of reasons. in job interviews. the embedded imperativeindexes a host of unmarkedRO sets." most find it COMPELLING. other directives are markedandevoke negative or perversereactionsfrom workers. this is a frequent (unmarked) directive form when participantsare unfamiliaror differ in rank. in effect.On a recentNationalPublic Radio interviewbemoaningthe behavior of job candidatesfrom today's twenty-somethinggeneration.that is. They do so knowing thata markedlinguistic choice will be viewed as a negotiationof some RO set OTHER thanthe unmarked one. speakersDO makemarkedchoices at times. which are operationally defined by theirfrequency(cf. By doing RO so. (1)Nevertheless. the embedded imperative with or without mitigators (e. (i) Recall. speakers are "agreeing"to "businessas usual". just as there is no universally unmarkedRO set. Thus. The candidatedidn't get the job. they accept the prevailing community views for an appropriate set. such as that between a salesperson and customerin many establishments. there is no universally unmarkeddirective.whetherfamiliarsor not. thus.g. Wouldyou please do x. Myers-Scotton 1998:31 for a fuller discussion of why unmarkedchoices prevail). as well as other situationalfactors. As Ervin-Tripp1976 pointed out. Speakersmake markedchoices as negotiationsto change the CURRENT RO set to a NEW RO set . that the unmarkedRO set varies across settings.when the task is outside the addressee's role.). given who the parRO ticipants are. even for work settings in the same community. although speakersmake choices as individuals. etc. For example.For example.g. for interactions in the auto factory studied by Bernsten 1998. most candidateswho expect to be successful observeandfollow not only the employer'sdemeanorbutalso his/ her linguistic taken as an information question.the one for which the CURRENT 10 Language in Society 30:1 (2001) . however.given the potentialcost of not but acknowledging its power. They make the more unmarkedchoices. speakers most frequently select the linguistic choice thatindexes what they perceive to be the moreunmarked set. the bald imperative indexes the unmarkedRO set between foremanand worker.for example. Consider. can you do x. the unmarkedchoice is often the one that meets the beliefs and desires of persons in the community who have sufficient power to set norms. As such. or when territorialityis at issue. (j) Inmost publicinteractions arepotentiallystatus-raising that (e. they typically behave as group members in that most make the same or similar linguistic choices.

and evaluating that course of action by imagining it. rather. but Damasio states that they are "the brain's representationof the body" (1996:1413). This is what the stories were telling us.what to expect next. In summarizing his interviews with fire departmentgroundcommanders. Klein (1998) proposes a RECOGNITION-PRIMEDDECISION model for explaining decisions in the life-threateningsituations he studied. In turn. I had been so fixated on what they were not doing that I had missed the real finding: thatthe commanderscould come up with a good course of action from the start. When a negative somatic marker is juxta- posed to a particularfutureoutcome the combinationfunctions as an alarmbell.Damasio's (1996:1413) hypothesis is thatsomatic markersenable the organismto take shortcutsin decision-makingthat have to do with "survival"at a numberof levels. The two primarysources of power in Klein's model are PATTERN RECOGNITION andMENTAL SIMULATION. Those emotions and feelings have been connected. actual physical experience is the input to these processes as well. It was not that the commanderswere REFUSING to compareoptions. Klein (1998:31) stresses the role of what he calls "intuition":"Intuitiondepends on the use of experience to recognize key patternsthat indicatethe dynamicsof the situation. Klein may emphasize the primacy of experience more than Damasio does. It seems clear thatDamasio's hypothesis that somatic markers Language in Society 30:1 (2001) 11 . Initially.the basis for action in Klein's "warstories"(some in the literal sense) seems very differentfromthe "bioregulatory processes"of Damasio (1996. italics in original). he writes that "somatic markers are a special instance offeelings generatedfrom secondary emotions.they did not HAVE to compare options. we will see how speakersemploy switching as a markedchoice.The exact nature of these markersis somewhat elusive. but both models clearly depend on (largely) unconscious cognitive calculations. The view that exposure to language in use sets up a sense of "readings of markedness"in the markednessevaluator corresponds to some of the ideas of Klein and also of Damasio. it becomes a beacon of incentive" (1994:173. by learning."Recognition includes realizing what types of goals make sense. Even when faced with a complex situation. Elsewhere. which cues are important.8 However.the processes permit organismsto set up SOMATICMARKERS. In short.CALCULATING SPEAKERS would be the unmarkedindex."Both are cognitively based components which help aid actors make judgments by utilizing experience. When a positive somatic markeris juxtaposed instead. In the codeswitching examples illustratedhere. and what course of action is likely to succeed.This model fuses two processes: sizing up the situation to recognize which course of action makes sense.Klein writes: MARKED choice Our results turnedout to be fairly clear. the commanderscould see it as familiar and know how to react. to predicted future outcomes of certain scenarios. (1998:17) We see Klein's "intuition"as not very different from the MM's "markedness evaluator. 1998).

the MM recognizes the constrainingand compellingnatureof norms. rememberthat speakersare always free to make markedchoices. legal and psychological constraintsthat an individual faces" (1989:14).not only on their compelling nature. or uncontested. they promote unmarkedchoices. As such. lest we be misinterpreted. rely on the notion of INTENTIONALITY in human actions. As Lessig states. All these models. Elster's structural constraintsare the social context and situational 12 Language in Society 30:1 (2001) . but also on their dynamic quality. the two filters in Elster'smodel of decision-making become three. One can even arguethat having successful interpersonalinteractionsis a form of "survival."He refers to structural constraintsas "all the physical.9 Lessig's (1995) contributionto this analysis is the emphasishe places on what he calls SOCIAL MEANINGS (which we interpretas derivedfrom norms). In the recastingprocess. they area backdropfor interpretingthe choices of others.. They assist the deliberationby highlighting some options (either dangerousor favorable). includingthe MM."Socialmeaningsact to induceactionsin accordancewith social norms. Yet he goes on to say. such architectures also be seen can 1 to coordinatereadings of intentionality. First. withoutbelieving thatthere is a single. However."something that Damasio himself suggests. Damasio writes.and THEREBY impose costs on efforts to transformsocial norms"(1995:998).or uniformacross any collection of people" (1995:954).' Finally.. For example. or stable."As for their effect. The institutions of society figure in Elster's first filter as "structuralconstraints. Recast in linguistic terms. and. "Somaticmarkersdo not deliberate for us. and meaningmanagement.whetheror not the individualchooses the power or constraints"(1995:955). Think of it as a biasing device" (1994:174)..I suggest. actors intend their actions to reflect goals or the very least. and eliminating them rapidly from subsequentconsideration. In interpersonal contexts.They empoweror constrainindividuals. 1 in terms of the MM's new filter model.CAROL MYERS-SCOTTON & AGNES BOLONYAI exist and his hypothesis about how they work are not very different from the MM's hypothesis thatspeakerstake accountof a markednessevaluatorin making linguistic choices. they do. economic.Underthe filtermodel presentedbelow..Elster's view of rationalityfigures prominentlyin our recastingof the MM. They also give at least a nod toward the notion that innate architecturescoordinate readings of cost-benefit analyses of competing choices. normsfigure as partof the availableevidence thata speakerconsiders in making choices. Following is an analysis of ex.the MM recognizes the very existence of normsand social meanings. "Butwe can speak of social meaning. agreed upon point for any social act. Nor of course are these meanings fixed.. as the examples will well as theirdynamicquality. Lessig says of social meaningsthat"notall meaningsareeasily recognized . Second.lo The MM echoes such views in two ways. Lessig crucially notes that the effects of social meanings are "in an importantsense nonoptional. and observers attributeintentions to actions.

but they are internalto the speaker. The third and final filter provides a rationale for his strategic use of switching. these elements are not under the speaker's direct control. however. 1. after he has switched to English in line 10.e." We equate this with the speaker's linguistic repertoire. Elster moves directly from the filter consisting of structuralconstraints to rationalityas a mechanism. Thus. The heart of Elster's RA model (his second filter. unmarked ways in which certain parts of conversation are realized. structurallydeterminedopportunityset. when he switches to English in lines 8 and 10 (I'll make my own salad). She can provide details about Krist6f's experience with linguistic choices in and outside the family. Based on this second filter. and then back to English again in line 12. This filter produces what Elster calls an "opportunityset. we add constraints that are more linguistic . What these two elements do is bias the selection of alternativesfrom the initial. we have the advantage that one of the speakers. is an analyst. the message the markednessevaluator gives him is that switching to English is a vulnerable move. it is an indirect refusal of her attempt to serve him salad. macroLanguage in Society 30:1 (2001) 13 . the mother. Ex. we add a new second filter at this point. now the third filter for the extendedMM) is where choice is located. Models of CS based on situationalfactors would predictthat Krist6f would speak English about school matters. Moreover.Note thatRC theories do not addresshow particularrepertoireshappento be available to certain individuals. Kristof's prior experience with language use in the family gives him evidence that his mother prefers him to speak Hungarianto her. To these. (He switches first back to Hungarian in lines 10-11. it contrasts with his mother's code choice. and (ii) the MM's markednessevaluator. and (iii) in its content.In our recastingof the filter model. Kristof's linguistic repertoireincludes English and Hungarian. please). Again. The biasing is in terms of previous experience.. We emphasize again thatchoice comes AFTER structuralconstraints have had their effects.g. this does not mean that such information is unimportant. Ilyen kicsi tdnyerokbancsinaljak a restaurant-okban'They make it in such small plates at the restaurants'. In an RC model. 1.only that it is outside the scope of the model. 1 offers strong support for the notion that such structuralconstraints as topic do not determinecode choices but only contributeto the speaker's repertoire. (ii) sequentially. This filter includes two innately available architectures which are filled in largely by experience. he knows that his choice is marked from his mother's point of view. Indeed. and thereforein terms of previous "successes" and "failures.instead." In assessing how this filter is relevant to ex. I need some salad. it is marked at a number of interrelatedlevels: (i) it goes against the unmarked RO set that holds in his interactions with his mother. he speaks Hungarianon this topic.CALCULATING SPEAKERS factors. All these constraints are external to the speaker. such as openings and closings.In ex.They are (i) the somatic markers of Damasio's 1996 model that influence the types of responses an organism has to stimuli. the subsequent switches seem to create confusion.

How does a markedswitch to English bringaboutoptimaleffects for Kristof?When he says I'll make my own salad (line 8 and again in line 10).family code. 1 in termsof the threeoperationsof rationalchoice. subordinaterecipientof his mother'sactions ( directs actors to perform these three operationsto find the best action: (i) actors consider their desires and values as well as priorbeliefs. This thirdfilter includes the mechanismsresponsible for the actual choice of an alternative.figure in promoting new norms). yet failing to observe the interactional norms(i. he will have a chance at successfully negotiating more independencein his actions. The third filter also includes social norms. andbeliefs take accountof availableevidence.. Myers-Scotton 1998. and (iii) finally.As he switches to English. HIS preferredlanguage. If he succeeds in negotiatingan RO set based on these images. 1999 for more details on how normsaffect linguistic choices andhow choices. values. lead him to this estimate: If he speaks English.just thatthey should be aware of any discrepancies and their consequences (cf. His desires and beliefs.Rationalityfunctions both as a mechanism and as an explanation. English is the language of the largercommunity. he disidentifies himself with the RO set that ascribes to him the role of a dependent.CAROL MYERS-SCOTTON & AGNES BOLONYAI societal featuresdeterminethe opportunityset. since nothing has changed in the context. the relathe tionshipamong(i) the choice of Hungarian. they make sure that their final desires. he accomplishes two things throughthe choice of English. as well as the values he attaches to each language and potential RO sets. In addition. English becomes a tool to exploit social meanings to his own ends.and (iii) the child's own view of his identity features is neither static nor exclusive. ingroup. Of course. As an explanation. Similarly. they take account of how these values jibe with the dictates of social norms. rationalitytells us WHY choices are made. (ii) the power relationsbetween the childrenandthe parent.The main element in this filter is rationality. Let us look at ex. he can conjureup in his mother'sview a new and "right" persona(a personaof the power andprestige associated with English). we might ask what good it does Kristof to switch immediatelyback to Hungarian(lines 10-1 1). (ii) they confirm that these three elements are internally consistent. In this new set. In view of the payoffs of speaking English.As a mechanism.e. but not how selections are made from the set. he calls for a competingRO set. switching to the less preferredlanguage)is certainlyone way in which Kristof can fight a battle over power.a CA explanationbased on preferenceorgani14 Language in Society 30:1 (2001) . This switch cannot possibly be attributedto contextual/situational factors. being servedthe salad). Second.compliantlittle boy.This does not mean they must forsake their own values . First.e. because one aspect of "available evidence" is the norms that apply and their social meanings. when speakers scrutinize (generally unconsciously) their values. he can display that he is about to pull out of a situationthat rendershim a passive. and the language associated with his social identity as a maturingboy who interacts with peers outside the home. in turn.

and beliefs and take account of available evidence). ANALYSIS OF OTHER EXAMPLES Two more examples of exchanges between Krist6f and his mother will now be analyzed. 8 K: (Spilling the water on the kitchen cabinet counter)Ah! Sorry.And making optimal use of the resources in his linguistic repertoire has its rewards: He gets to make his own salad. but it has a different ending.' Language in Society 30:1 (2001) 15 . against his mother's wishes. note that. Bocsdnat. my God.As Elsterputs it. Ex. AND his mother even gives him more vegetables. Let me just do it by myself. because a sequentialapproachprovides no mechanismfor the co-occurrenceof a dispreferred(English?) and a preferred(Hungarian?)language choice within one turn.' 4 K: Csindlj limonddet. "Itmust only be well groundedin the available information"(1994:23).e. 'Make a lemonade.hogy inni akarsz. Interpretations about intentions offered for ex. ne keljfel. 'I know that you want to have a drink. In summary. Within a rationality-basedframework. sorry! 9 K: Nagyon nehe'zvolt ez. 1 and the following examples are not held up for verification.this sub-model of three filters. Clearly. majd en adok. is posited in orderto explain code choices.It mitigates the weight of the face threat of his noncompliantbehavior. don't get up. is a message of deference towardhis mother.however.A rationalbelief does not have to be objective or true. rationality is not to be confused with objectivity or truth.' 3 M: Tudom. check the internal consistency of desires. 1. 'I'll make a lemonade.the model presumes that much of what "happens"is below the surface: Speakers' intentions surface as code choices.' 2 K: Limonddetcsindlok. I'm sorry. I'll get you one. I know how much 6 sugar I want in the lemonade. 7 K: (Lifting a bottle of water) Oh.even though he returns to English in line 12. with rationality at its heart. and. 1 M: Vdrjcsak. values.CALCULATING SPEAKERS zation also falls short. However. 'Wait. they have no empirical basis. as an explanation. followed by three examples from other settings and other language pairs. 2 is similar to ex. Kristof sets out to make his own lemonade: (2) Making lemonade.' 5 K: (Switch to English) I'm gonna get the ingredients. At the same time. The family is at the dinnertable again. Speaking Hungarian. we claim that these interpretationsare not simply one analyst's readings of the data. we can claim that Kristof's switches are internally consistent and motivated by the goal to optimize his returns. 'It was too heavy. The test that they have some measure of independence from the analyst is the extent to which they are consistent with the rational mechanism that RC theory claims actors use (i.

'Dad said that it was here. switching back to Hungarianhas its costs: He cannot maintainthe image of an independentperson.By "returning the fold" of Hungarian. he knows he can reinstatehimself within the RO set preferredby his mother if he speaks Hungarian. switching FINALLY to Hungarianhas some fringe benefits other than just restoringharmonywith his mother. this switch back into the unmarkedcode for family dinner-tabletalk comes as a result of unconscious but rational calculation.Krist6fhas multiple identities. 'I don't know. 'I'll look for it later. 'It was too heavy.' 2 M: Majd megkeresem. god. you need to change clothes. 'Top. continues between motherand Hanna) 32 K: (to himself) Oh.he apologizes first in English and then in Hungarian. his switch NOW to Hungarianis a MARKED choice. The accidentcauses him to lose face.' 3 K: Ah. But the switch also has its rewards:Krist6f makes sure his mother understandsthis second apology as a "pledge of allegiance" to reinstate himself within the confines of the RO set thathis motherprefers. his apologies show a reversepattern.CAROL MYERS-SCOTTON & AGNES BOLONYAI When Kristofspills the water(lines 7-8). the language associated with his outgroup identity. Within a single turn.' 4 M: Kristdf. "Switch to a code that expresses deference to others when special respect is called for by circumstances"(Myers-Scotton 1993:147 ). In ex. too. 16 Language in Society 30:1 (2001) .' (conv. You need to look there. 'Ah. His mother and his five-year-old sister Hanna (H) are at the dinnertable in the dining room. 'Krist6f. In addition. you're so mean.('Ah. 'In a bag.' (Conversationcontinues between mother and Hanna.First eat your dinner.Meg kell ne'zniott is.' 14 K: Fent vagy lent? 'Top or bottom?' 15 H: Fent. Krist6f is looking for his hat in the hallway closet so that he can go outside and play.Of course. de mean vagy.) 12 K: (looking for the hat in the closet) Papa azt mondta. governed by the deference maxim of the MM. hogy te megkeresed. and his to self-image as an autonomousboy is damaged. sorry.Again.hogy itt van. In this respect.Like any otheractor. 3. Bocsanat.' 16 M: Nem tudom. he can at least gain face by refurbishinghis image as a compliant son. 'Where's my hat? Dad said that you would look for it.Based on experience.Edd meg el6bb a vacsorddat. Recall that the accident happens when he is speaking English. sorry!' in English is followed by Nagyon nehe'ezvolt ez.' 13 M: Egy zacskoban. English symbolizes independence from parentalcontrolfor him. His understandingof the norms and values of language use in relation to his mother enter into this equation. I'm sorry').dt kell oltoznod. (3) Where's the hat? 1 K: Hol van a sapkam?Azt mondtaa papa.

its double meaning arises from the fact that he uses the marked language to convey a literally innocent and polite request in order to slip in his real feelings.ethnographicmilieu. as his frustration grows. followed by his more confrontationalremarkin line 41. (0. he resorts to sarcasm.CALCULATING 33 34 35 H: K: SPEAKERS 36 M: 37 K: 38 M: 39 40 41 42 43 44 K: M: K: M: K: M: Mi van? 'What is it?' It fell.' Mi? 'What?' Nincs a zacsk6ban. This success may give parity"with his mother that. line 32) and when answering his sister (It fell. Abban a szekre'nyben. CS examples from other settings and other language pairs provide further insights into how speakers'choices are driven by rationalityratherthan derived directly from social groupmembership. ott lehet. He switches to English when talking to himself (Oh. igaz.e.' Krist6f. Kristof's use of English (i. I'm asking you very-very kindly. It's not here. you see. The language of the conversation between Kristof and his mother is Hungarian. However. 'Well. hogy a kocsiban van? 'Mama?Isn't it possible that it's in the car?' Hdt lehet. the speakture. he can always hide behind the "outlaw"mask it provides ("It's not really me who's speaking"). line 34).. 'It's not in the bag. it can only be there.. each speaker Language in Society 30:1 (2001) 17 . 'In that closet. nowhere else. it is possible. 'Then where is it?' Tesse'k? 'Pardon?' Where is it? Please tell me. or sequentialstruccouple. in line 43. The utterance is obviously ambiguous. 'Krist6f. Kristof switches to English (line 39) for a remark targeted at his mother. Speaking English may grant him less accountability for his remark.' Mama?Nem lehet.In a conversationin Berlin between anAmerican-German ers use non-converging discourse in the unmarkedRO set (i. God. mdshol nem. Krist6f is anxiously going through the bags in the closet in search of his cap.' Following his mother's instruction.e.' Akkormeg hol van? I'm asking you very-very kindly.8) It's not here. he him a sense of enough "interactional can suggest (albeit in Hungarian)that his mother is wrong about where the cap must be. Whereis it? Please tell me) seems to "succeed" in the sense that his mother does not censure him. Constrained by the norms of politeness as well as the need for his mother's help. Since English is not associated with the unmarkedRO set for this interaction. ldtod. that's true. 'It's in the bag! It's in the bag! It's not in the bag. A zacskoban van! A zacsk6ban van! Nincs a zacsk6ban.the unmarkedcode. switching to the marked code for his saucy remark in line 39.

Can you lend it to me?' H: Du kriegst eine neue Brille jetzt. you know what I had. meine Augen sind schlecter. I got my eyes checkedOh. So my right eye has gotten much worse. eigentlich. 'But you see. 'Yeah. certainly.speaks only passable English. in the last few years. is quite fluent in German. 'account' H: Ja. 'I orderedthem. were you at the-' S: Brillenwerkstatt. Du warst bei der'Where were you.whereI got these. ich renne blind durch die Gegend! 'Yeah.ja. And um. what I did. Right? 'lenses' H: M-hmm S: And it's the same man that I know from 1990.' S: No. new Glaser. especially the right eye. Du sagst Du siehst immerso viel mit der Brille. what should I do. the American.while the young man Hans (H).CAROL MYERS-SCOTTON & AGNES BOLONYAI speaks his/her native language). Fuller 1997 discusses the example in terms of calculations of accommodation. he said. 'Yes. on your Konto.' H: Uh-huh. er sagte.' H: Tatsachlich? 'Really?' S: Ja.' S: Do you have that? I thoughtyou had nothing in. ja. now. das kostet 300 Mark mit Krankenversicherung. klar. Und ich mufi300 Mark bezahlen. Hans. wo ich die habe. 1 S: 2 H: 3 S: 4 5 6 7 8 9 10 11 12 13 14 Oh. whereI got them. yeah. I'm running around blind. 'The Brillenwerkstatt. 'Of course. I'll take it from my savings account. 1989. And um.And um. Hans. actually. yeah. they've gotten worse. ich nehme das von meinemSparkonto. or what?' S: Ja.' Language in Society 30:1 (2001) 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 18 . the German. can you loan me 300 marks?Do you have it? H: Natiirlich.' (Discussion continues in German) S: Aber das Problem ist. where? -and my eyes are worse. Kannstdu mir ausleihen? 'But the problem is. H: Wowarst Du. (4) Money for glasses (Fuller 1997:74-76). was soll ich tun. The young woman Sarah(S). I need 300 marksnext Monday.' H: Aber Du siehst. I need a new.And I have to pay 300 marks. you say you see so much with those glasses. S: Ich habe die bestellt. my eyes have gotten worse. that costs 300 marks with health insurance.wo ich meine Brille. oder was? 'You're getting a new pair of glasses. ich brauche 300 MarknachtstenMontag.

but Sarah still persists in English in her next turn. but then she switches to Hans's unmarkedchoice. Her words (That'swhat I wanna know. Now eleven lines later. Sarahswitches back to English. to introduce the main topic: a needed sum of money. in line 52. All your other comments are unnecessary. German. For Sarah. her unmarkedchoice in their conversations. Sarah sees that being compliant andjustificatory in Germanhas not helped." Furthermore. she probes further(Geht das? Ist das keinProblem? 'Will thatwork?Is thata problem?'). she switches back to English to confront Hans directly about the needed money.' S: Okay. or by means of Bell's Language in Society 30:1 (2001) 19 . she is assuredby Hans (Kein Problem 'No problem') that he will lend her the money. the whole point of the conversationis to get Hans to help her pay for the pair of glasses that she needs. or even in whatever might count as "contextualeffects. Sarah begins in her native language. How does this switch negotiate maximumutility for her? Her willingness to speak Germanshows deference to Hans. In terms of an RC model. to find out where the money will come from. Earlier (line 41) Sarah had asked for a loan as part of her deferential switch to German. It would be difficult to explain Sarah'smany switches as motivated by either a change in anything that could be called situationalfeatures or structuralorganization. English. based on available evidence. and at the same time she can make sure that he clearly understandswhy she needs money. That's what I wanna know. Giles & Coupland 1991). Fuller points out how Sarah uses a marked choice. Having reached her goal. She tells Hans about having had her eyes checked. Finally."she returnsto her own native language. e. Hans's English is only passable. although some of her choices can be explained in an accommodationframework (cf. Not quite satisfied with his reply. Hans replies immediately in the affirmative.CALCULATING 57 58 59 60 S: SPEAKERS Geht das? Ist das kein Problem? 'Will that work? Is that a problem?' H: Kein Problem. This and subsequentswitches show how she makes a series of choices to optimize her own outcome. She switches to English and at the same time escalates herrequestin the form of two questions (Can you loan me 300 marks? Do you have it?). as she asks for money.but she has now switched to German.g.All your other commentsare unnecessary) make it clear thatthis switch to English means that the accommodationis over. she cannot rely on it. Finally. which in this case is to converge to monolingual Germandiscourse with Hans. Note thather first switch to Germancomes in the turnin which she first mentions the issue of money (lines 10-14). Giles and associates. and how much. 'No problem.perhapsas a mitigation.when it has become clear to her that accommodatingto Hans by speaking Germandid not "help. She maintainsGerman throughwhat must have been an annoyingly involved explanation of what she was doing (orderingglasses) and why.

mother (A) and son (B) have different preferences for unmarked languages in mother-son interactions. These approachescan explain SOME of every speaker's considerations. it may be in construedas a means to projecta powerful and distancedpersonafor the mother. language preferences differ across generations.After a brief delay.British-bornchildrennormallypreferto speak English. the mother asks the son in English if he has finished his homework. 20 Language in Society 30:1 (2001) .this exchange opens up to a slightly differentinterpretation. these approachesdo not work for all of Sarah'schoices. he respondsin English.Her switches that are very "scattered"at the surface level are easily explained within an RC framework as not "scattered"at all." which promptsthe mother to reformulateher question in Cantonese. but as parts of a consistent. mother'sinitial choice of English is unexpected. Viewed in terms of an RC framework. whereasparents normally use Chinese in intra-generationalinteractions (Li 1995). Either generation may switch to the other's preferredlanguage "for certain communicative effect" (Li & Milroy 1995:296).however. It is particularlythe productionof indeterminacyof readings by these ambiguous choices that most clearly defies a single-minded or inductive form-to-functionutteranceinterpretationin sequentialanalyses. they seem to pay less attentionto why and how speakersmake choices thatrepresentexceptions to the unmarkedpreferencestructure. Li & Milroy provide an excellent analysis of "whatis going on" at the local level of turns.However. In many conversations in this community. Next. this choice evokes associations of English as "thelanguageof authority" the Tynesidecommunity(Li 1995).Her The motive seems to be to suspend the unmarkedRO set in order to bring about a desired effect. On the other hand. Sarah'sswitches are prime examples of language use based on cognitive calculationsto optimize outcomes. I've finished. The 12-year-oldboy is playing with a computerin the living room as his mother is trying to find out if her son did his homework. As a markedchoice.but it is hardto see how they can continuously orient to a speaker's changing calculations-just what seems to be going on with Sarah.On the one hand. thus. because English is her son's preferredlanguagebut not hers.The mother prefers to speak Cantonese. Thus.he rightly points out that the boy's answercan also be seen as a tying back to the mother'sfirst question in English. the mother's use of English can be considered deferential. the mother's speaking of English is loaded with ambiguity. goal-directed strategythatends up requiringmuch switching. His answer is a "noticeable silence. and her son prefers English. consider an example from the Tyneside Chinese community in Britain.CAROL MYERS-SCOTTON & AGNES BOLONYAI 1999 recasting of his audience design model to highlight speakers' choices as initiatives as well as responses to audience. In ex. Li 1995 explains the boy's contrastive language use as a "dispreferredsecond part"to his mother'sindirectrequestto study. First. One characteristic featureof languageuse on which an RC model can shed light is speakers' ability to employ marked choices to pursue a complex set of goals and maintainmultiple role relationships. 5.

Her authorityis recognized. Li 1995:205). yiu mo wan sue? 'Wantto review (your) lessons?' 4 B: (1. ineffective) choice of English to her unmarkedchoice of Cantonese. (6) Daughterand cakes (Halmari 1993:17). the motheruses Finnish for each offer of cake. Note that this pattern is just the opposite of what a CA-based analysis Language in Society 30:1 (2001) 21 .A cost-benefit analysis tells her to switch to Cantonese. 1 A: Finished homework? 2 B: (2. It's better. whereas in line 4. otaksaa tata kakkua? 'Irene.' 3 M: Ota sitten tosta toi . Setting:The mother(M) offers two differenttypes of cake to the younger child (S2). Under an RC interpretation. Although Halmariexamines how the children (aged 8 and 9) use CS as an evaluative device. in this case.The switch displays how the motherweighs and prioritizesher goals differentlyat this point: Takingaccountof the availableevidence (lack of response to English).' 4 S2: Yes. (5) Getting a response from a 12-year-oldboy playing with the computer(Li & Milroy 1995:288. Halmaripoints out that it is importantto note that while S2 is still a fluent bilingual. she uses English for the acceptance. the data also provide further insights into how speakers' choices are driven by rationalityratherthan by sequential structure.CALCULATING SPEAKERS Note that when the motherdoes not get an adequateresponse to her question (Finished homework?). main goal is to have her son pay attentionand her respond.yesterday's pie. 1 M: Irene. Presumably. she satisfies her ultimategoal by switching to Cantonese. Thus. In lines 1 and 3. she shows a preferencefor English. The switch is from the marked(and.she has the choice of continuing in English or switching to Cantonese. she makes her goal and her preferredRO set (i. and the son replies.Recall thatunderan RC approach assessing availableevidence is one of the mechanismsthatguide choices. do you want this cake?' 2 S2: Emmdatykkdacheesecakesta 'I don't like cheesecake.0) 3 A: Steven. the mother'schoice to switch to Cantoneseis motivatedby the evidence thatboth mother and son know that she has some measure of authorityin the RO set that her use of Cantonese indexes. the girl uses Finnish for the refusal. The final example (6) comes from two Finnish/English bilingual sisters living in the United States (Halmari 1993). this approach leads to quite a different interpretationthan just that the mother turns to Cantonese for a contrastive effect. 'Take then that . albeit in English. In line 2. The focus is now on HER desires and goals and not on accommodatingto her son in any way. When she does not succeed in establishing her authorityabout school mattersthroughspeaking some English.5) I've finished.eilinen leivos. her desires and values) unambiguousfor her son.e.

although speakers do not necessarilyhave "cooperation" theirprimaryinterpersonal as objective. They are also filtered through internalconstraints.rationality. in both its content (It's better) and its form (language choice contrastingwith that of the addressee. the child presents herself as deferentialtowardher mother. but her own preferredchoice).In line 2. RO sets.As a "reward. in CA parlance . and so they communicatewith the assumptionthat their conversationalcontributionwill be available to others for interpretation. choices in a rationally basedmodel of linguisticvariationpass through several filters. Finally. pay attentionto. evidence is STORED . since the language of refusal .the innatelyavailablearchitectures markednessevaluator. the child's respective language choices show how she finds the "goldenmean"of accomplishingtwo interactionalgoals and conveying two types of social messages simultaneously. It may be more or less transparent thus may requiredifferent degrees and types of inferenand tial work.which in turn are constrainedby large-scale societal factors and the discourse structureof their communities. speakers operate with the expectation that what they say will be available evidence for others. Within an RC model. values. the girl's self-assertive the centerpiece. aims at restoring interactionalparity.not a new code. and even as somatic intuitions. (a somatic markers)that bias choices based on experience. In addition. as a consequence of the use of Finnish for the refusal. In addition. Evidence is what can be seen or heard. They begin with the external constraintson speakers: their linguistic repertoires. choices pass througha thirdfilter in which a social mechanism. CONCLUSION In summary. the rejection is mitigated. they always are acutely aware of their listeners. and that they assess these in regardto internalconsistency and available evidence. speakers collect. a preferredsecond part. is done with no code alignment. However. 22 Language in Society 30:1 (2001) . they speak to be heard.shows code-alignment.The focus from the other shifts to the self as the girl projects a more powerful self-image groundedin her own desires and value system. certainly as norms. her language of acceptancedoes not show deference any longer. frames. Also at odds with a CA analysis is the fact that acceptance of the offer. as rational actors. In addition.Evidence is often inextricably linked to the beliefs and values that also guide actors: the stuff of beliefs and values is recycled evidence.CAROL MYERS-SCOTTON & AGNES BOLONYAI would predict. To act rationally means that speakerstake account of their own beliefs. and take account of all these sources of available evidence in calculating the possible outcomes of their decisions regarding HOW to speak.Simply stated. And. and goals." is offered she the cake that she likes. Perhaps taking account of available evidence is the most fundamentalfeature of rational action.a dispreferredsecond. Indeed.

Three limitations of this frameworkas a model of linguistic choices (and limitationsof RC theoryin general)areworthemphasizing. in whateverways are in importantto them and are rationally grounded. inequality in socio-economic opportunities. along with rationalityas an explanatorymechanism.but speakers.rationality"at work"are more conspicuously manifest in Westerninteractionsis that. RC theory is based on the assumptionabout human cognition that actors are oriented to seek optimality of an interpersonalnaturein their actions. A specific goal of this essay has been to show how the MM. Some might claim that rationality is a culturally biased. In effect. Rationality has to do with what goes on in the mind of the individual (how an individual makes choices) and not with individualism (a value). with some recasting. the Markedness Model (MM) is an RC model. speakerssee these social meanings AS A RESOURCEfor making choices. explains these conversationswith a betterfit between the model's premises andthe datathando two othermajorapproachesto linguistic variation. given the actor's beliefs and values. We disagree. Western notion.cultural differences in beliefs and values (the impetus for optimizing outcomes) are what makes rational acts appeardifferent cross-culturally.not the presence or lack of rationality per se in a culture. The same can be said for the tacit knowledge thatspeakershave aboutpreferredorganizationalpatternsin conversation. as a rationallybased informationthatis externalto the speaker impinges on and/or shapes the speaker's set of opportunities. We have analyzed six conversations that include switches of languages to what can be considered markedchoices within the terms of the interactionat hand. The overall assumptionis thatthe way speakerschoose to speak reflects their cognitive calculations to present a specific persona that will give them the best "return" theirinteractionswith others. Although speakersare certainly awareof salient social factors in theircommunities thatresult in associationsof social meaningswith particular linguistic varieties (whether these factors are dominantideologies. and we have employed the basic premises of the MM. or otherlarge-scale societal factors). generally speaking.CALCULATING SPEAKERS The speaker's beliefs and values merit more comment. leading to a set of plausible hypotheses abouthow courses of action. the gap between individualistic and group values is wider there than in collectivist cultures. including linguistic choices. The reason why individualpreferencesand desires .which we have called a social factors model and ConversationAnalysis (CA).Therefore. Note that an RC theory says that an action is rational if it appears to optimize outcomes.not opportunities.12RC theory is based on these deductivepremises. including theirlinguistic choices. One goal of this article has been to show how.First.make decisions. mechanisms Language in Society 30:1 (2001) 23 . can be explained. not as a determinantof choices. such a theory assumes thatthe way to explain social phenomenais in terms of rationalityas the social mechanism employed by individuals in their decisions and manifested in their behavior.

g. Reims. 1 For example. analysts) might consider to be rationallybased choices. and that these choices are rationally based . such models do not claim that actors always make what others (e. for example. The closeness of the fit between the premises and the data makes such a model attractive. quoting from Heritage (1984:242): of A speaker's action is context-shapedin that its contributionto an on-going sequence of actions cannot adequatelybe understoodexcept by reference to the context .. especially. education. a rationality-based potency. Meeuwis & Blommaert 1994:418. Grice introduceshis well-known exposition of the Cooperative Principle and its maxims with this comment:"As one of my avowed aims is to see talking as a special case or variety of purposive.given what one knew (and could have known)at the time" (1997:761.paperswhose goals are to correlatelinguistic data with demographicfeatures.such as codeswitching. italics in original).As Elster puts it. but do not necessarily predict.A further source of attractivenessof the fit is thatthe explanationsit offers aremore general in a universal sense than are the social and structuralprofiles of the individual phenomenato be explained. France. 4 In their introductionto the volume RethinkingContext. 2 Witness. and gender. future choices individualswill make. it may be worth noting that the specific expectations or presumptions connected with at least some of the foregoing maxims have their analogues in the sphere of transactionsthat are not talk exchanges" (1975:47). and Steven Gross for comments on an earlier version. Its strengthlies IN THE FIT between two deductivepremises. this approachinsists thatit is the place of the individual'ssocial group vis-a-vis other social groups that crucially affects choices.. such as socio-economic status.including.This contextualizationof utterancesis a major and unavoidableprocedurewhich hearersuse and rely on to interpretconversational contributionsand it is also something which speakers pervasively attend to in the design of what they say.CAROL MYERS-SCOTTON & AGNES BOLONYAI such as rationality allow us to explain. NOTES *This is a much revised version of a paper given in July 1998 at the InternationalPragmatics Conference.Janet Fuller. .Duranti& Goodwin (1992:28-9) offer this characterization CONTEXT. behavior. For example. 3 In its more extreme form. model of linguistic choices still has a special Nonetheless. the immediatelyprecedingconfigurationof actions in which it participates. indeed rational. our subjecthere cannot be explained "withoutexplaining what society does to speakersand vice versa"(cf.To be rationaldoes not mean that one is invariablysuccessful in realizing one's aims: it means only that one has no reason to think that one should have acted differently. the many paperson social dialects in AmericanEnglish at NWAV(New Ways of Analyzing Variation)conferences every year . some would argue that the individual'smotivationsfor even micro-level interactions. nor do they claim that actors always make the best choices from an objective standpoint. Third. in a negative review of Myers-Scotton 1993).13 Second. "Rationalchoice models are subjectivethroughand through. 24 Language in Society 30:1 (2001) . Note that even the title of this review ("The 'MarkednessModel' and the absence of society") reflects the authors'attitudethat not giving primary attention to large-scale societal factors is a fatal flaw in any attempt to explain codeswitching.and data-basedanalyses relying on these premises. an RC model does not necessarily produce quantitativeevidence.thatspeakers seek optimality when making choices. We thank Roberta Chase-Borgatti.

8 Damasio makes very clear the connection between mind and body: When this process [decision-making]is overt. whichever it is.externalor internalon the speaker's choice are evident."The rational/utilitarian tradition.They argue that language users employ discourse connectives (because they convey proceduralknowledge ratherthanconceptualknowledge) to constrainthe interpretation readings. the device ence.see Ben-Rafael 1993.A related problem "is that most things people make decisions about are not measured in money. at least) in his second chapter. This approachis popular. that itself is something which gets displayed in the next turnin the sequence. When the process is covert the somatic state constitutes a biasing signal. That understandingmay turn out to be what the prior speaker intended. his discussion of efforts to overcome these problems focuses on social policies. for instance through a non-specific neurotransmitter influences cognitive processing. "There are limits on the ability to process information and make rational decisions" (1994:153). a speaker's choice may indeed be "anact of identity. On the group level. (1998:40) 9Althoughsuch a system as thatproposedunderthe somaticmarkerhypothesiswould have evolved to maximize basic survival. Hudsonendorses it in the second edition of his text. something to which a and "dispreferred" sCA refers to "preferred" speaker attends before making a contribution. Language in Society 30:1 (2001) 25 . Sociolinguistics (1996). Operatingin a cognitively based framework.and therefore from a very different point of view.However. It is plausible that a system geared to producemarkersand signposts to guide basic survival. though these ideas are attractive. nevertheless areon the forefrontin attemptingto apply sociological insights to propose policies that have a realistic chance of succeeding" (1994:178). The somatic state is alerting you to the goodness or badness of a certain optionoutcome pair. For the view of a sociologist of language on theories of rationality. the problem of how social solidarity is created requires better solutions: "There is the question of how rational individuals can form groups: indeed. sociolinguists can gain insights from the proposed rational solutions to create group solidarity. We can agree with Hudson's rationale "thateach individual is unique [and social space"(1996:29). are indirectly linked to precisely the same framework of survival versus danger. of what follows.for all theirfaults. giving rise to the idea that "pure"rationality cannot exist. and hence are not strictly comparable"(1994:155).He concludes: "Themodernrational/utilitarians. it would seem that positing knowledge of what is preferredimplies that speakers take account of "something"in addition to the immediate context."He points out several immediately obvious problems with RC approaches. Using an indirect and non-conscious influsystem such as dopamine. of gain and balance versus loss and disequilibrium. That is. in Wilson & Sperber 1993 and Blakemore 1992 discuss what might be called a form of "preference" how discourse is organized. constraints. on the individual level.First.they are hard to apply. (1996:1417) Damasio points out that the somatic markerhypothesis "developed as a response to a numberof intriguingobservationsmadein neurologicalpatientswith focal damagein the frontallobe" (1998:36). Hutchby& Wooffitt (1998:15) discuss context in this way: "Speakersdisplay in their sequentially 'next' turns an understandingof what the 'prior' turn was about. or not.CALCULATING SPEAKERS In a more recent exposition of ConversationAnalysis.for example. only "bounded"rationality. Since Collins is a sociologist. the somatic state operates as an alarm signal or an incentive signal."but it is not clear whetherthereis any principleddeliberationimpelling no a speakerto make one choice ratherthan another. Collins (1994:121-180) gives an excellent critical overview of the history and present state of Rational Choice theory (within sociology. Collins notes. Damasio goes on to say: A very large range of other problems including those which pertain to the social realm." parts of adjacency pairs. that]individualsuse languageso as to locate themselvesin a multi-dimensional However. would have been pre-adaptedto assist with "intellectual"decision making. in effect setting up "preferred" 6 The "acts of identity" approachof LePage & Tabouret-Keller1985 seems to be a type of Rational Choice Model because speakersare viewed as pursuingtheirown goals in the ways they choose to speak. of advantage versus disadvantage. still.Furthermore. how collective action is possible at all" (1994:153). The device produces its result at the openly cognitive level.

"It's really trying to maximize an expected value keeping in mind the multiplicity of possibilities. see Nuyts 1993 for an argumentagainst the claims made by these authors.CAROL MYERS-SCOTTON & AGNES BOLONYAI These patientslargely retaintheir intellectual abilities. We are well awarethat not all researchersaccept the idea that intentionalityis a universalfeature of language use. (1987). "Summersis always weighing probabilities"(1998: 38). Styling the other to define the self: A study in New Zealand identity marking. Bing. 10 Elster offers us this gem about norms: "To paraphrase Weber. Codes and consequences: Choosing linguistic varieties.. Charles (1998). Integratingcooperationand conflict: Commentson RaymondBoudon's paper. Peter (1998). Bergvall. reason. Allan (1999). for all practicalpurposes. Journal of Sociolinguistics 3:523-41.a newly appointedSecretary of the Treasury. Randall (1994). New York: Grosset/Putnam. In a featurestory on LawrenceSummers. Four sociological traditions. Philosophical Transactionsof the Royal Society. the long-termoutcome for a type of response option).Jacob Weisbergwrites. Rethinkinglanguage and gender research: Theoryand practice. and violence in an infanticide case. "The inferences seem determinate.a social norm is not like a taxi from which one can disembarkat will" (1994:31). Eliezer (1993)."adding. Blakemore.Janice (1998). Descartes' error: Emotion. JanetM. a memory in fact. The somatic markerhypothesis and the possible functions of the pre-frontalcortex.. He goes on to quote Summers as saying. Bernsten. He writes.). eds. but they show problemsin theirpersonallife. Antonio (1994). REFERENCES Auer. Ben-Rafael. Code-switchingin conversation. The extraordinary thing is thatit seems. New York:Oxford University Press. Oxford:Blackwell. Levinson refers to one of the kinds of human intelligence as INTERACTIONAL INTELLIGENCE (1995:222) and to "the astounding speed of conversation inferences [as] something that should be noted" (1995:239). Diane (1992). between the disposition for a certain aspect of a situation(for instance. (1996). & Freed." Weisbergquotes Summersfurther:"Thatbasically is the canonical model for makingrationaldecisions.Alice F. Bell.InternationalStudies in the Philosophy of Science 7:29-3 1. 178-91. II That speakers are predisposedto make readings of intentionalityis one of the main points of Levinson 1995. Damasio goes on: "I propose that the ventromedialprefrontalcortex establishes a simple linkage. and thus communication. Stephen C. 26 Language in Society 30:1 (2001) . Briggs. areall people who have implicitly made sensible calculationsbased on the fact that there are a multiplicity of possibilities. Damasio. "Human interaction. Cf. 12 That humanchoices are rationallybased is a view not only of academics but also of crucially placed governmentalofficials. sexuality.[are] socially inadequate and are demonstrablydifferent from the choices they were known to have made in the premorbid period"(1998:36). Cambridge:CambridgeUniversity Press. Brown. London B 351:1313-1420. ed. In Carol MyersScotton (ed. London:Routledge. However. Notes on a "Confession":On the constructionof gender. and the disposition for the type of emotion that in past experience has been associated with the situation"(1998:39). and the human brain. Politeness: Some universals of language usage. people who make personalchoices well. VictoriaL. Penelope & Levinson. People who makejob choices well in their lives. who cite argumentsby a numberof anthropological linguists thatthis idea aboutintentionalityis only a Westernnotion. Understandingutterances. Myers-Scotton (1993:153-4) gives some indication of the predictions one might make aboutthe social profiles of the types of personswho will make more markedchoices thanothers and those persons who will maintainthe status quo throughunmarkedchoices. In referenceto the specific structuresinvolved. New York:Oxford University Press. depends on intentionascription. Collins. London:Longman." 13 However. (1996). "The choices these patients made are no longer personally work most of the time" (1995:241). Pragmatics 7:519-46.though we are happy to revise them when forced to do so. Meeuwis & Blommaert 1994. Markedversus unmarkedchoices on the auto factoryfloor.

W. & L. October 1998.Jan (1994). Duranti. Grice.). Fishman. (1990). Interactionbiases in human thinking.In Ralph Fasold & Deborah Schiffrin (eds. (1976). Paperpresented at NWAVEConference. P. Conversationanalysis. Talkat work. Code-switching.Alessandro. preference markingand politeness in bilingual cross-generational talk: Examples from a Chinese community in Britain. La Linguistique 2: 67-88.Susan. Klein. Andr6e(1985). Lessig. Roberts. Penelope (1989). 1-57. Stephen C.Nikolas (1991). Eckert. University of Georgia. Acts of identity. The intersectionof sex and social class in the course of linguistic change.Ian. Logic and conversation. Language: Contextsand consequences. Guerra. & Tabouret-Keller.William (1966). Language in Society 21:1-26. The somatic markerhypothesis and the possible functions of the prefrontalcortex. Who speaks what language to whom and when. Cambridge:Cambridge University Press. City. Sources of power: How people make decisions. Exact descriptionof the speech community:ShortA in Philadelphia. Frame analysis.DC: Center Labov.& Coupland.and ChantalTetreault(eds. Ulysses and the sirens. Howard. (1997). & Blommaert. Hutchby. Social network and social class: Toward an integrated sociolinguistic model. Paul. University of Chicago Law Review 64:749-64.In PeterCole & JerryMorgan(eds. Cambridge:CambridgeUniversity Press. sity Press. Anne-MarieP. Social intelligence and interaction. Variationas social practice. Richard(1996). Synthese 98: 21-49. T.emotions.Washington. Code-switching as an evaluative device in bilingual discourse. Oxford: Oxford University Press. Charles (1992). Lawrence (1995). Accountingfor Tastes.CALCULATING SPEAKERS (1998). Sociolinguistics. Hedstrom. In Esther Goody (ed. Proceedings of the Fourth Annual Symposiumabout Language and Society (SALSA). Cambridge:CambridgeUniversity Press. Erving (1974). Halmari. & Heritage. Elster. Meeuwis. Li. Becker (1996). Journal of Multilingual & Multicultural Development 16:197-214. The "MarkednessModel" and the absence of society: Remarkson codeswitching. (1975). John (1992). Language Variationand Change 2:205-5 1. Conversationalcode-switching in a Chinese community in Britain: A sequential analysis. Cambridge:CambridgeUniversity Press. Issues in Applied Linguistics 4:91-118.Cambridge:CambridgeUniverLePage. More than enough. Language in Society 30:1 (2001) 27 . H. The cement of society. Amsterdam:Benjamins. Co-constructingbilingualism:Non-convergingdiscourse as an unmarkedchoice. In A. Journal of Pragmatics 23:281-99. Pacific Grove. (1998). C. New York:TeachersCollege Press. 221-59.Peter.MA: MIT Press. Rationality. Social mechanisms:An introductoryessay. New York:Academic Press. 1-31. Heritage. Levinson. Joshua (1965). Goffman. Robbins. Jocks and burnouts:Social categories and identityin the high school. Richard(1998). New York:Harper& Row. & Milroy.Janet(1997). 41-58. Drew. 2nd ed.). Language change and variation. University of Chicago Law Review 62:944-1045. (1989). Social mechanisms. Conformistand non-conformist:Resolving the contradictionin women's use of language. (1994). Fuller. & Goodwin. "Is Sybil there?"The structureof some American English directives. 36-80. Cambridge:Polity Press.). Robin (1998). Oxford: Blackwell.).TexasLinguistic Forum 37:68-77. (1999). Gary (1998). The regulation of social meaning. Language in Society 5:25-66. James (1995). In Peter Hedstrom& RichardSwedberg (eds. Thesocial stratificationof English in New York for Applied Linguistics. Lesley. Cambridge:CambridgeUniversity Press. (1995). Milroy. eds. and social norms. Jon (1979). Hudson. Rethinkingcontext. Weiskrantz(eds. (1989).).Cambridge:Polity Press. Multilingua 13:387-423. CA: Brooks/Cole. Li.). Syntaxand semantics. Robert. Cambridge:CambridgeUniversity Press.Lesley (1995). Giles. Wei (1995). & Milroy. Garfinkeland ethnomethodology. In Alice Chu. Review of Gary S. & Swedberg. The prefrontal cortex. Michael. Ervin-Tripp. & Wooffitt. John (1984). eds. Wei. Helena (1993). Cambridge.

(1998). Linguistic form and relevance. 2nd ed. Schilling-Estes. A theoretical introductionto the MarkednessModel.Carol (1983). and Sperber. American Speech 75:3-33. & Wilson. Lingua 90:1-25.Dan.Dan (1993). Deirdre. (1993). Intentionsand language use.Karl R. The negotiation of identities in conversation:A theory of markedness and code choice. 40. Popper.& Beckett. Natalie (1998).Walt. The logic of scientific discovery. New York: Oxford University Press. InternationalJournal of the Sociology of Language 44:115-36. Investigating "self-conscious" speech: The performanceregister in OkracokeEnglish.Jacob (1998). 53-55. Wolfram.). Jan (1993). 38. Wilson. Explainingthe role of normsand rationalityin codeswitching. Nuyts. Weisberg. Keeping the boom from busting. Language in Society 27:53-83. & Sperber. Journal of Pragmatics 32:1259-1271. Oxford: Blackwell.CAROL MYERS-SCOTTON & AGNES BOLONYAI Myers-Scotton. Relevance: Communication cognition. (1959). 18-38. In Carol Myers-Scotton (ed. New York TimesMagazine 147:24-29. 73. The role of the individualand group in earlierAfricanAmerican English. (1999). Codes and consequences: Choosing linguistic varieties. revision accepted 21 January2000) 28 Language in Society 30:1 (2001) . Deirdre(1995). Dan (1999). (Received 17 October 1999. Antwerp:AntwerpPapersin Linguistics. New York:Basic Books. Oxford:ClarendonPress for Oxford University Press. Social motivationsforcodeswitching:EvidencefromAfrica.